Build binary using juliac

[michel2323]

@amontoison @frapac I've played a bit around with juliac for🎄 from this talk.

I've managed to build a binary of ExaPF with only CPU support. It seems to me that CUDA/GPUCompiler.jl generated code cannot yet be digested through juliac, even if you don't trim. I've tried multiple things. Maybe there is a way.

$ juliac ExaPFApp --output-exe ExaPFBin --bundle exapfdir --experimental
◑ Compiling...LLVM ERROR: Cannot select: intrinsic %llvm.nvvm.membar.cta

[107726] signal 6 (-6): Aborted
in expression starting at none:0
Allocations: 797069608 (Pool: 797066510; Big: 3098); GC: 383
✓ Compiling...
ERROR: Failed to compile ExaPFApp
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:44
 [2] compile_products(recipe::JuliaC.ImageRecipe)
   @ JuliaC ~/.julia/packages/JuliaC/0OTq8/src/compiling.jl:123
 [3] _main_cli(args::Vector{String}; io::Base.TTY)
   @ JuliaC ~/.julia/packages/JuliaC/0OTq8/src/JuliaC.jl:191
 [4] _main_cli(args::Vector{String})
   @ JuliaC ~/.julia/packages/JuliaC/0OTq8/src/JuliaC.jl:182
 [5] main(ARGS::Vector{String})
   @ JuliaC ~/.julia/packages/JuliaC/0OTq8/src/JuliaC.jl:197
 [6] _start()
   @ Base ./client.jl:556

This hit a CUDA intrinsic that gets sent to the Julia (CPU) LLVM backend.
I've added a README.md if you want to start playing too.

[frapac]

Anton has tried to compile MadNLP with juliac, and has some trouble to get GPU support working. See the latest developments in : https://github.com/apozharski/libmad

Cross-ref: JuliaGPU/GPUCompiler.jl#611

[michel2323]

Thank you, @frapac, for the cross-ref! I was looking and looking... (skill issue). Seeing that the right people are on it, I'll just continue with the toolchain and do a CPU-only release, assuming this will be fixed sometime soon.

[amontoison]

@michel2323 I tested it on kkt 👍

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 70.86%. Comparing base (9277944) to head (9164068).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #310      +/-   ##
==========================================
+ Coverage   70.75%   70.86%   +0.10%     
==========================================
  Files          29       29              
  Lines        3925     3871      -54     
==========================================
- Hits         2777     2743      -34     
+ Misses       1148     1128      -20     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.